Prediction and sensitivity analysis of embankment dam settlement under earthquake loading using gene expression programming

Abstract

ABSTRACT A correct understanding of dams’ dynamic behaviour can help improve structures’ design and safety. The settlement of earth dams caused by earthquake shaking can induce instability and destruction. Using white-box machine learning techniques, including gene expression programming (GEP), a numerical relationship has been tried to estimate the settlement of earth dams under different types of earthquake loading, so that this relationship can be used in the design of earth dams. The main input parameters used in this research include the ratio of the yield acceleration of the dam to the maximum horizontal acceleration of the earthquake (ay/amax), the ratio of the fundamental period of the dam to the predominant period of the earthquake (Td/Tp), the magnitude of the earthquake (Mw) and the settlement ratio (S/H) earth dam as the output parameter. The R, MAE, and RMSE values for the proposed relationship in the research were equal to 0.85, 0.058, and 0.127, respectively. The measurement of statistical parameters shows the acceptable accuracy of the GEP method in predicting S/H. However, developing a database containing other effective parameters can be considered for future research.